Author(s)

A.R. Elobeid Ahmed

Abstract

An Arabic phoneme recognition system using Hidden Markov Models (HMM) is introduced. This system is an important step towards the realization of a continuous speech recognition system with a large size of Arabic vocabulary. A discrete HMM is implemented for modeling each of the Arabic phonemes. Training and recognition are both based on Viterbi methods. For deciding on the best features that can represent Arabic speech signals, performance tests were implemented on a number of parametric representations such as prediction coefficients, area function, cepstral coefficients, etc. Results showed the superiorit

Keywords